HBERTv1_48_L2_H768_A12_ffn_2 / train_results.json
gokuls's picture
End of training
52dcaee
{
"epoch": 6.36,
"train_loss": 3.953961916252199,
"train_runtime": 197999.1228,
"train_samples": 5858758,
"train_samples_per_second": 2958.982,
"train_steps_per_second": 26.9
}